CAN-QTree: A Distributed Spatial Index for Peer-to-Peer Networks

Abstract

We present a distributed spatial index called CANQTree based on a Content-Addressable Network (CAN) andQuadTree-alike structures. In our system, both spatial objects and their corresponding nodes in a QuadTree are identified by some points and mapped into CAN zones of peers. For a given spatial range query, CAN-QTree can provide O(n 1/2 ) search performance. With a uniform distribution of spatial objects, given ε > 0, CAN-QTree can provably maintain a load imbalance of at most 2 + ε between a highest loaded peer and a lightest loaded peer. Simulations show that our CAN-QTree is an effective spatial index with a guarantee of good load-balancing.

Low-Cost Search in Tree-Structured P2P Overlays: The Null-Balance Benefit

“…One challenge in tree topologies is balancing their height. Tree topologies such as m-LIGHT [18], CAN-QTree [22], P-Grid [1], and the m-ary tree [8] may fail in balancing the tree height. These approaches cannot guarantee a bounded cost for their search operation.…”

“…[19] forms an indexing architecture where a zone is assigned to each cluster in a Zigzag hierarchy. CAN-QTree [22] is a tree extension of the Content Addressable Network (CAN) [13]. It uniforms a distribution of spatial objects with a search performance of O(n 1/2 ).…”